AetherSoft unveils lightweight on-device LLM and raises $40M Series B

AI · 5 min read

AetherSoft's WhisperLite is a compact LLM distilled for short-form assistant tasks like summarization, intent routing, and mnemonic recall on constrained hardware. The model is quantized and packaged with an optimized runtime that supports heterogeneous mobile NPUs and microcontrollers.

The company raised $40 million in Series B funding led by Silicon Ridge Partners to expand model family support, improve quantization toolchains, and add privacy-preserving training methods. AetherSoft plans to deliver SDKs for major mobile OSes and partner with OEMs to preinstall the runtime.

AetherSoft positions WhisperLite for scenarios where users want offline assistants with low latency and guaranteed data residency. Pilot integrations include a note-taking app that offers offline summarization and a camera app that tags scenes without uploading images to the cloud.